I’ve been playing with Adobe GIL, following it’s inclusion into Boost, and I’ve been getting some fun compile errors in relation to <windows.h>. Note that, even if you aren’t including <windows.h>, some of the boost headers do.
Here lies: What not to do, so you don’t waste hours on it like I did.
Test-case:
int main()
{
::gil::rgb8_image_t a;
::gil::jpeg_read_and_convert_image("a", a);
}
Compiled with Visual Studio 2005 SP1, Boost 1.33.1 and JPEG 6b (27-Mar-1998):
cl /EHsc jpegio.cpp /I c:\sdk\libs /I c:\boost\include\boost-1_33_1
First attempt, <windows.h> first:
#include <windows.h>
#include <gil/core/gil_all.hpp>
#include <gil/extension/io/jpeg_dynamic_io.hpp>
Error:
jpegio.cpp
c:\sdk\libs\gil\core\channel.hpp(90) : warning C4003: not enough actual paramete rs for macro 'min'
c:\sdk\libs\gil\core\channel.hpp(92) : warning C4003: not enough actual paramete rs for macro 'max'
c:\sdk\libs\jmorecfg.h(161) : error C2371: 'INT32' : redefinition; different bas ic types
C:\Program Files\Microsoft Visual Studio 8\VC\PlatformSDK\include\basetsd.h(62) : see declaration of 'INT32'
c:\sdk\libs\jmorecfg.h(215) : warning C4005: 'FAR' : macro redefinition
C:\Program Files\Microsoft Visual Studio 8\VC\PlatformSDK\include\windef.h(145) : see previous definition of 'FAR'
Second attempt, other way around, Error:
jpegio.cpp
C:\Program Files\Microsoft Visual Studio 8\VC\PlatformSDK\include\basetsd.h(62): error C2371: 'INT32' : redefinition; different basic types
c:\sdk\libs\jmorecfg.h(161) : see declaration of 'INT32'
Third attempt:
#include <windows.h>
#include <gil/core/gil_all.hpp>
#define XMD_H
#include <gil/extension/io/jpeg_dynamic_io.hpp>
#undef XMD_H
This should, in theory improve matters, as XMD_H being #defined stops <jmorecfg.h> trying to redeclare INT32.
Oh no.
Error:
jpegio.cpp
c:\sdk\libs\gil\core\channel.hpp(90) : warning C4003: not enough actual parameters for macro 'min'
c:\sdk\libs\gil\core\channel.hpp(92) : warning C4003: not enough actual parameters for macro 'max'
c:\sdk\libs\jmorecfg.h(215) : warning C4005: 'FAR' : macro redefinition
C:\Program Files\Microsoft Visual Studio 8\VC\PlatformSDK\include\windef.h(145) : see previous definition of 'FAR'
c:\sdk\libs\gil\core\color_convert.hpp(148) : error C2589: '(' : illegal token on right side of '::'
c:\sdk\libs\gil\core\color_convert.hpp(221) : see reference to function template instantiation 'void gil::color_converter_default_impl<lt;T1,C1,T2,C2>::operator ()<lt;SrcP,DstP>(const P1 &,P2 &) const' being compiled
with
[
T1=SrcChannel,
C1=SrcColorSpace,
T2=DstChannel,
C2=DstColorSpace,
SrcP=gil::pixel<lt;gil::bits8,gil::cmyk_t>,
DstP=gil::pixel<lt;gil::bits8,gil::rgb_t>,
P1=gil::pixel<lt;gil::bits8,gil::cmyk_t>,
P2=gil::pixel<lt;gil::bits8,gil::rgb_t>
]
c:\sdk\libs\gil\extension\dynamic_image\../../core/image_view_factory.hpp(71) : see reference to function template instantiation 'void gil::default_color_converter::operator ()<lt;gil::pixel<lt;T,C>,gil::pixel<lt;T,gil::rgb_t>>(const SrcP &,DstP &) const' being compiled
with
[
T=gil::bits8,
C=gil::cmyk_t,
SrcP=gil::pixel<lt;gil::bits8,gil::cmyk_t>,
DstP=gil::pixel<lt;gil::bits8,gil::rgb_t>
]
c:\sdk\libs\gil\extension\dynamic_image\../../core/image_view_factory.hpp(69) : while compiling class template member function 'gil::pixel<lt;T,C> gil::color_convert_deref_fn<lt;SrcConstRefP,DstP,CC>::operator ()(gil::pixel<lt;T,gil::cmyk_t> &) const'
with
[
T=gil::bits8,
C=gil::rgb_t,
SrcConstRefP=gil::cmyk8_ref_t,
DstP=gil::pixel<lt;gil::bits8,gil::rgb_t>,
CC=gil::default_color_converter
]
c:\sdk\libs\gil\extension\io\jpeg_io_private.hpp(166) : see reference to class template instantiation 'gil::color_convert_deref_fn<lt;SrcConstRefP,DstP,CC>' being compiled
with
[
SrcConstRefP=gil::cmyk8_ref_t,
DstP=gil::pixel<lt;gil::bits8,gil::rgb_t>,
CC=gil::default_color_converter
]
c:\sdk\libs\gil\extension\io\jpeg_io_private.hpp(178) : see reference to function template instantiation 'void gil::detail::jpeg_reader_color_convert<lt;CC>::apply<lt;V2>(const View &)' being compiled
with
[
CC=gil::default_color_converter,
V2=gil::rgb8_view_t,
View=gil::rgb8_view_t
]
c:\sdk\libs\gil\extension\io\jpeg_io.hpp(147) : see reference to function template instantiation 'void gil::detail::jpeg_reader_color_convert<lt;CC>::read_image<lt;Image>(Image &)' being compiled
with
[
CC=gil::default_color_converter,
Image=gil::rgb8_image_t
]
jpegio.cpp(12) : see reference to function template instantiation 'void gil::jpeg_read_and_convert_image<lt;gil::rgb8_image_t>(const char *,Image &)' being compiled
with
[
Image=gil::rgb8_image_t
]
c:\sdk\libs\gil\core\color_convert.hpp(148) : error C2143: syntax error : missing ')' before '::'
c:\sdk\libs\gil\core\color_convert.hpp(148) : error C2780: 'T gil::channel_invert(T)' : expects 1 arguments - 0 provided
c:\sdk\libs\gil\core\channel.hpp(235) : see declaration of 'gil::channel_invert'
c:\sdk\libs\gil\core\color_convert.hpp(148) : error C2059: syntax error : ')'
c:\sdk\libs\gil\core\color_convert.hpp(149) : error C2589: '(' : illegal token on right side of '::'
c:\sdk\libs\gil\core\color_convert.hpp(149) : error C2143: syntax error : missing ')' before '::'
c:\sdk\libs\gil\core\color_convert.hpp(149) : error C2780: 'T gil::channel_invert(T)' : expects 1 arguments - 0 provided
c:\sdk\libs\gil\core\channel.hpp(235) : see declaration of 'gil::channel_invert'
c:\sdk\libs\gil\core\color_convert.hpp(149) : error C2059: syntax error : ')'
c:\sdk\libs\gil\core\color_convert.hpp(150) : error C2589: '(' : illegal token on right side of '::'
c:\sdk\libs\gil\core\color_convert.hpp(150) : error C2143: syntax error : missing ')' before '::'
c:\sdk\libs\gil\core\color_convert.hpp(150) : error C2780: 'T gil::channel_invert(T)' : expects 1 arguments - 0 provided
c:\sdk\libs\gil\core\channel.hpp(235) : see declaration of 'gil::channel_invert'
c:\sdk\libs\gil\core\color_convert.hpp(150) : error C2059: syntax error : ')'
…
Fourth attempt, the “right” way:
#include <gil/core/gil_all.hpp>
#define XMD_H
#include <gil/extension/io/jpeg_dynamic_io.hpp>
#undef XMD_H
#include <windows.h>
This works as intended.
Solutions?
Assuming that we’re not going to get the JPEG code patched, the INT32 definition from <windows.h> comes from <basetsd.h>, which declares the BASETSD_H header-guard macro.. it’d be nice if GIL issued an explanation of what was going to go horribly wrong, instead of that lovely template error in test three?
